[OpendTect_Users] Seismic Volume Alignment using MatchDelta / DeltaResample

[Friso Brouwer]

Hello Colin,

Yes we apply this attribute often in house, it was originally designed for
aligning partial azimuth stacks, but should also be usable for partial
offset stacks.

Just a few ideas:
On improving the delta cube, I guess partial offset stack may be more
different between themselves than partial azimuth stacks:

   - The near and far may have different frequency content, you may want to
   analyze that, and apply a high-cut to the near, before feeding it to the
   delta calculation.
   - The near and far may have zones where they have very different
   reflectivity, for example where you have HC induced phase reversals and I
   normally these are the zone you want to get it right most of all. I guess
   that is where you want to use the cross-correlation based filter(?) In
   addition maybe there are some attributes you can use as input for the delta
   calculation that would suffer less from the difference in near and far
   reflectivity that could be input to the delta calculation, such as
   (combinations of) phase or absolute amplitude attributes (I am guessing
   here, but I would say it is worth testing).


On filtering the delta cube:

   - Making a cross-correlation attribute/cube:
      - Create a mathematics attribute, near-far
      - Use the volume statistics attribute to create a time windowed rms
      attributes for (near-far) attribute, near and far volumes.
      - The use the mathematics attribute to calculated 1-(rms{near-far} /
      (rms{near} +rms{far} ) ) as cross correlation attribute (this is
OpendTect's
      similarity equation, of course there are other cross-correlation
equations
      you can use instead).



   - For the conditional filter, use the if then else statement in the
   mathematics attribute, this would look like this:

   (crosscorrelation_attribute < c0) ? replace delta with something : keep
   delta
   , with c0 a user defined threshold



   - of based on cross correlation and ve value:
   ((crosscorrelation_attribute < c0) || (|ve|>c1)) ? replace delta with
   something : keep delta
   , withc0 and c1 user defined thressholds.

I would advise for the 'something" a median filtered version of the delta,
created by a time windowed volume statistics

Of course you can make these equations again more sophisticated, and apply
gradual conditions.

Please let me(us) know what worked and what not.

Kind regards,

Friso

On Thu, May 19, 2011 at 3:38 PM, Colin Hawke <colinhawke at hotmail.co.uk>wrote:

>  Hello,
>
> Has anyone successfully used the Match-Delta & Delta-Resample to align two
> seismic volumes? I have a near & far angle stacks that I'd like to align.
> However the Delta cube (time-shift volume) that is created by MatchDelta is
> very noisy and so when applied using DeltaResample the output is not at all
> desirable.
>
> What would make sense to me is to filter the time-shift volume before
> applying it. The best way I can think of would be to filter out samples that
> had a low cross-correlation between near & far. Even better might be to do
> it two-dimensionally and filter out samples that had large +ve or -ve
> time-shifts AND a low cross-correlation values. In any case I can't seem to
> figure a way to do the above work flow.
>
> Another method could be to apply dip-steered median/averaging filtering to
> the time-shift volume to smooth things out but that seems less than ideal.
>
> What methods or workflows are people using out there?
>
> Kind Regards
>
> Colin Hawke
>
> Predict Geoconsultants Ltd
>
> _______________________________________________
> Users mailing list
> Users at opendtect.org
> http://lists.opendtect.org/mailman/listinfo/users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opendtect.org/pipermail/users/attachments/20110520/1442515e/attachment.html>